code-review

name	code-review
description	Use when: reviewing GitHub PR via gh CLI, scale-adaptive strategy
allowed-tools	Bash, Read, Grep, Glob
triggers	["code-review","/code-review","review PR","PR review"]

Code Review: Scale-Adaptive GitHub PR Review

Role Detection (READ THIS FIRST -- MANDATORY)

Check your initial prompt. If it contains the literal string "Use the code-review skill", then:

YOU ARE THE EXECUTOR. Follow these rules:

1. SKIP the "Execution Protocol" section below -- it is for the orchestrator, not you.
2. Read the pattern at ../../PATTERN.md relative to this SKILL.md, and follow it step by step.
3. ABSOLUTE PROHIBITION: Do NOT run csa run, csa review, csa debate, or ANY csa command. You must perform the work DIRECTLY. Running any csa command causes infinite recursion.

Only if you are the main agent (Claude Code / human user):

• You are the orchestrator. Follow the "Execution Protocol" steps below.

---

Purpose

Perform AI-powered code review on GitHub PRs with automatic scale adaptation. Small PRs (< 200 lines) are reviewed directly; large PRs (> 800 lines) are delegated to CSA. Includes authorship detection to use adversarial debate for self-authored code, and mandatory AGENTS.md compliance checking for all changed files.

Execution Protocol (ORCHESTRATOR ONLY)

Prerequisites

• PR number must be known or discoverable

Quick Start

csa run --sa-mode true --skill code-review "Review PR #<number>"

SA Mode Propagation (MANDATORY)

When operating under SA mode (e.g., dispatched by /sa or any autonomous workflow), ALL csa invocations MUST include --sa-mode true. This includes csa run, csa review, csa debate, and any other execution commands. Omitting --sa-mode at root depth causes a hard error; passing false when the caller is in SA mode breaks prompt-guard propagation.

Step-by-Step

1. Fetch PR context: Use gh pr view and gh pr diff --stat to get metadata and diff statistics.
2. Assess scale: Small (< 200 lines) = direct review. Medium (200-800) = direct with progress. Large (> 800) = delegate to csa review.
3. Authorship check: If Co-Authored-By matches caller model family, use csa debate instead of direct review.
4. For large PRs: Checkout PR branch locally, run csa review --branch <base>. Do NOT pre-read diff.
5. For small/medium PRs: Fetch full diff via gh pr diff, analyze each file for quality, security, performance, and language-specific issues.
6. AGENTS.md compliance: Discover AGENTS.md chain for each changed file. Check every applicable rule. Violations are P2+; MUST/CRITICAL rules are P1.
7. Generate review: Produce structured output (summary, critical issues, suggestions, nitpicks, questions, AGENTS.md checklist).
8. Post review (optional): Submit as PR comment via gh pr comment only when user explicitly requests.

Example Usage

Command	Effect
/code-review 42	Review PR #42 with auto scale detection
/code-review 42 post=true	Review and post comment to the PR
/code-review scope=base:main	Review current branch changes vs main

Integration

• May use: csa-review (for large PRs), debate (for self-authored code arbitration)
• Used by: pr-bot (as part of PR review loop)

Done Criteria

1. PR metadata and diff statistics fetched successfully.
2. Scale-appropriate review strategy selected and executed.
3. If large PR: delegated to CSA review (not read into main agent context).
4. If self-authored: adversarial debate review used instead of standard review.
5. AGENTS.md compliance checklist generated with all applicable rules checked.
6. Structured review output produced (summary, issues, suggestions, checklist).
7. If post=true: review comment posted to PR via gh pr comment.